Posts from machine learning

Empire AI

Last summer I sat down with Tom Secunda, who co-founded Bloomberg LP with Mike Bloomberg, to talk about areas of shared philanthropic interest. Tom told me that academic institutions do not have access to the kind of AI/ML infrastructure that the top tech companies have and he wanted to fix that. His idea was a consortium of Universities in New York State, the New York State Government, and philanthropic donors. His vision was a large shared facility in upstate NY with state-of-the-art AI/ML infrastructure that participating academic institutions could make available to their faculty for cutting-edge AI/ML research.

Tom is a convincing person. He convinced me that this was a good idea last summer and he went on to convince Governor Hochul and the top Universities in New York State and his fellow philanthropist Jim Simons.

I am glad Governor Hochul and her team were quick to recognize the promise of this idea. Today, Governor Hochul will announce Empire AI in her State of the State Address.

Empire AI will be a “state-of-the-art artificial intelligence computing center in Upstate New York to be used by New York’s leading institutions to promote responsible research and development, create jobs, and unlock AI opportunities focused on public good.”

Over $400mm of public and private funding has been committed over ten years to build and operate the Empire AI facility. New York State is contributing $275mm and over $125mm is coming from participating Universities and philanthropy.

I am excited to see New York State step up like this. Other states, like Massachusetts, have done something similar but this NYS effort is significantly larger. I expect more states will follow now. Cutting-edge AI/ML research should not be limited to large tech companies. We need our academic institutions to be on equal footing. This model, particularly if more states adopt it, can help make that possible.

New York is one of the leading AI centers in the US, along with California and Massachusetts. We see this every day as entrepreneurs building AI companies come knocking on our door. It is very encouraging to see our local government supporting and investing in this new area of economic development.

I want to congratulate Governor Hochul, the leaders of our academic institutions, and Tom Secunda, for their vision and initiative here. This is important.

#machine learning#NYC#VC & Technology#Web/Tech

Software That You Can't Shut Down

The term “censorship resistant” is used a lot in the decentralized computing/web3/crypto space to talk about a core feature of these systems. I don’t love the term censorship resistant because it is a wonky term.

Software that is encoded in smart contracts (and other ways) on fully decentralized blockchains can’t be shut down or turned off.

So why is this a big deal?

Let’s make up a story:

Imagine that an AI is trained to teach children to read better than humans. But the powers that be decide that teaching children to read is something that humans need to do. So they make this AI illegal and block access to it.

Well if this AI is written in a smart contract on a fully decentralized blockchain, it can’t be shut down. As long as there are nodes somewhere around the world willing to maintain the blockchain, this AI will continue to run and anyone that has access to the Internet will be able to use it.

That’s a made-up story, but hopefully, you get the point. I picked AIs and education, but you could also go with self custody and your money. Or you could go with image detection and speeding cameras.

The point is that the powers that be have, from time to time, decided that certain things are bad and attempted to shut them down. Alcohol, for example. Or sex between two people who love each other, for example. Or books about racism, for example.

Fully decentralized blockchains offer something powerful. Software that can’t be turned off. Data that is open to everyone. AIs that can’t be shut down.

We are in a moment with enormous posibilities brought on by the computer science revolutions in machine learning, decentralized systems, and new user interfaces. It will be tempting for powerful entrenched interests to seek to put the genie back in the bottle on some of this stuff. But if the genie is deployed on a fully decentralized blockchain, there is no going back in the bottle.

#blockchain#crypto#machine learning#Web3

AI Art

There has been a lot of discussion about how AIs can make art and possibly replace artists, but I think the opposite is more likely to happen. Artists have been using AI to make art for a while now and the pace has picked up a lot in recent years.

I have always loved the work of Ian Cheng who makes computer-generated simulations that evolve using artificial intelligence. His works change infinitely. The first time I saw that, maybe ten years ago now, it made me rethink many ideas I had about art.

With the introduction of NFTs, artists can now make, release, and sell AI-generated art much more easily.

This week, our portfolio company Bright Moments has a big event in Tokyo, and one of the collections being shown features eleven top AI artists.

Though I could not make it to Tokyo this week, I was able to acquire a number of fantastic works in the collection.

My favorite is this piece by Claire Silver which is one of a series she calls paracosm.

Claire said this about the work:

This collection is a visualization of part of the artist’s paracosm. A text-to-image model was trained on some of their memories of that world and its inhabitants. 

I also quite like Helena Sarin‘s Kogei Kats. I picked up this one:

Helena’s website says that “Since 2021 her main creative energy is directed towards the #potteryGAN – making ceramics using her GAN/AI work as designed to 3D functional objects.” I really dig that.

I am very bullish on the creativity that AI will help artists bring to the world. It is a tool, like a paintbrush or a camera, or a kiln. And they will use that tool to make work that will bring great joy to our lives. They already are.

#art#machine learning

AIVC

I was approached by a company this week that has trained a large language model on all of the blog posts I have written here at AVC. There are 9059 of them for those that are counting. They wanted to offer me a chat bot called “ask Fred.”

I told them no thanks.

Let me explain.

I am totally fine with anyone using all of the content I have produced here at AVC to train their AI models. When I started AVC, I put a creative commons license on the content here. It has always been my view that anything I write here is in the public domain. You can repost it. You can do what you want with it. I just need attribution and a link back to the original post. That’s been my deal since the earliest days of AVC.

But an AI is not me. When you ask me something, you get my brain on the problem.

I have put a lot of what is in my brain onto the page here at AVC. But I have not put all of it.

I also don’t think an AI has my humanity, my ego, my empathy, my love, or my hate.

Maybe someday that will change. But we are not there yet. I think we are a long way from that.

So if you want to ask Fred something, you will still have to approach me.

#machine learning

Etsy Lens

I am the Chair of the Etsy Board and have been an investor and board member at Etsy since the mid-2000s. It is a company that I love and get great joy from being part of. Last year Etsy quietly launched a feature that has completely changed the way I use Etsy. It is called Etsy Lens.

Etsy has millions of items for sale in its marketplace but shopping on Etsy is generally not intent-driven. It is idea-driven. Most people don’t go to Etsy and enter “pizza oven” into the search field. A more common search would be “red pillow for my couch.” As a result, searching on Etsy can be a bit of a “hunt and peck” experience, even as the search on Etsy has improved enormously in the last few years.

I was in a coffee shop in a hotel in NYC this morning and saw an antique typewriter that I thought was great. I opened my Etsy app and got the search field.

I clicked on the camera icon and my phone took a photo of the antique typewriter:

I clicked the blue checkmark and Etsy gave me these search results:

I have been using Etsy so much differently since finding out about Etsy Lens. I see things that I like when I am out and about, use the Etsy app to photograph them, search Etsy for similar things, favorite and curate them in my profile, and then buy the ones I love.

When I showed Etsy Lens to the Gotham Gal, she said “Take photos of things you like to find things you will love.” That sums up Etsy Lens so well for me.

#machine learning#marketplaces

The AI Assist

I wrote last week that I have started coding again. And I have been amazed at how much easier it is now that I can code and deploy in the cloud without having to spin up anything myself.

But the other massive improvement in programming is the “AI assist.”

I am working in a Javascript library called jQuery and I don’t really understand its commands and syntax very well.

So I turned to ChatGPT last week and got back this:

That is like having every line of code commented out so you know exactly what it is doing. Once I understood what the code was doing, it was pretty simple to edit it to do something different.

GitHub also has a service called CoPilot that I have just set up so I haven’t used it yet. They call it “your AI pair programmer” which sounds like exactly what I need. I hope to get it working this week and that will help me even more.

Like all things AI, some will say that machines will replace humans in writing code. I think that could happen, but what certainly is happening is machines are making humans more productive in writing code. AND AI is allowing humans who aren’t very good at writing code to be able to do it much more easily.

The machines replacing humans narrative is powerful. But the narrative I prefer is that AI is making things available that have been expensive and unobtainable for so many. And that is not limited to programming. It is true of so many things.

#machine learning

Machine Learning and Schools

I read last week that the NYC Department of Education has banned ChatGPT from its networks and devices. I understand that reaction and mentioned the issues that AI/ML create for educators in a post a few weeks ago.

I attended a dinner this past week with USV portfolio founders and one who works in education told us that ChatGPT has effectively ended the essay as a way for teachers to assess student progress. It will be easier for a student to prompt ChatGPT to write the essay than to write it themselves.

However, I would like to suggest that educators embrace these new tools rather than block them.

We are entering an era when AIs will be available to everyone to use to do work, entertain ourselves, and many other things. We cannot put this genie back in the bottle. We need to embrace it.

I think a better approach would be to require students to use ChatGPT to write an essay or at least help write an essay and then have the students compete to see who can leverage this technology to create the best essay. That would teach the students to use these tools rather than pretend they don’t exist.

I own a slide rule that my dad gave me. He used it for many years until calculators emerged. He told me when calculators first showed up, many educators wanted their students to continue to use slide rules. But eventually, they realized that calculators were better and embraced them.

I think the same thing will happen with AIs. So we might as well get busy integrating them into education instead of banning them.

#hacking education#machine learning

Sign Everything

The advances in AI over the last year are mind-boggling. I attended a dinner this past week with USV portfolio founders and one who works in education told us that ChatGPT has effectively ended the essay as a way for teachers to assess student progress. It will be easier for a student to prompt ChatGPT to write the essay than to write it themselves.

It is not just language models that are making huge advances. AIs can produce incredible audio and video as well. I am certain that an AI can produce a podcast or video of me saying something I did not say and would not say. I haven’t seen it yet, but it is inevitable.

So what do we do about this world we are living in where content can be created by machines and ascribed to us?

I think we will need to sign everything to signify its validity. When I say sign, I am thinking cryptographically signed, like you sign a transaction in your web3 wallet.

I post my blogs at AVC.com and also at AVC.Mirror.xyz which is a web3 blogging platform that allows me to sign my posts and store them on-chain. This is an attestation at the end of last week’s blog post.

You can see that “author address” and click on it to see that it is one of the various web3 addresses I own/control. That signifies that it was me who posted the blog. It is also stored on-chain on the Arweave blockchain so that the content exists independently of the blogging platform. That is also important to me.

I think AI and Web3 are two sides of the same coin. As machines increasingly do the work that humans used to do, we will need tools to manage our identity and our humanity. Web3 is producing those tools and some of us are already using them to write, tweet/cast, make and collect art, and do a host of other things that machines can also do. Web3 will be the human place to do these things when machines start corrupting the traditional places we do/did these things.

#art#blockchain#bots#crypto#digital collectibles#hacking education#machine learning#non fungible tokens#streaming audio#VC & Technology#Web/Tech#Web3

What Happened In The 2010s

My friend Steve Kane suggested I take a longer view in my pair of year end posts this year:

And so I will.

Here are the big things that happened in tech, startups, business, and more in the decade that is ending today, in no particular order of importance.

1/ The emergence of the big four web/mobile monopolies; Apple, Google, Amazon, and Facebook. A decade ago, Google dominated search, Apple had a mega hit on their hand with the iPhone, Amazon was way ahead of everyone in e-commerce, and Facebook was emerging as the dominant social media platform. Today, these four companies own monopolies or duopolies in their core markets and are using the power of those market positions to extend their reach into tangential markets and beyond. Google continues to own a monopoly position in search in many parts of the world, has a duopoly position in mobile operating systems, and controls a number of other market leading assets (email, video, etc). Apple owns the other duopoly position in mobile operating systems. Amazon has amassed a dominant position in e-commerce in many parts of the world and has used that position to extend its reach into private label products, logistics, and cloud infrastructure. Facebook built and acquired its way into owning four of the most strategic social media properties in the world; Facebook, Instagram, Messenger, and WhatsApp. Most importantly, outside of China, these four companies own more data about what we do online and also control many of the important channels to reach us in the digital world. What society does about this situation stands as the most important issue in tech at the start of the 2020s.

2/ The massive experiment in using capital as a moat to build startups into sustainable businesses has now played out and we can call it a failure for the most part. Uber popularized this strategy and got very far with it, but sitting here at the end of the 2010s, Uber has not yet proven that it can build a profitable business, is struggling as a public company, and will need something more than capital to sustain its business. WeWork was a fast follower with this strategy and failed to get to the public markets and is undergoing a massive restructuring that will determine the fate of that business. Many other experiments with this model have failed or are failing right now. When I look back at the 2010s, I see a decade during which massive capital flowed into startups and much of it was wasted chasing the “capital as a moat” model.

3/ Machine learning finally came of age in the 2010s and is now table stakes for every tech company, large and small. Accumulating a data asset around your product and service and using sophisticated machine learning models to personalize and improve your product is not a nice to have. It is a must have. This ultimately benefits the three large cloud providers (Amazon, Google, Microsoft) who are providing much of the infrastructure to the tech industry to do this work at scale, which is how you must do it if you want to be competitive.

4/ Subscriptions became the second scaled business model for web and mobile businesses, following advertising which emerged at scale in the previous decade. Startups that developed the skills to execute a subscription business model with positive unit economics delivered fantastic returns to investors and capital flowed into this sector as a result. This was a very positive development as subscriptions better align the interests of the users and the developers of mobile and web applications and avoid many of the negative aspects of the free/ad supported business model. However, as we end the decade, a subscription overload backlash is emerging as many consumers have signed up for more subscriptions than they need and in some cases can afford.

5/ Silicon Valley’s position as mecca for tech and startups started to show signs of weakening in the 2010s, largely because of its massive successes this decade. It is incredibly expensive to live and work in the bay area and the quality of life/cost of life equation is not moving in the right direction. The physical infrastructure (transit, housing, etc) has not kept up with the needs of the region and there is no sign that it will change any time soon. This does not mean “Silicon Valley is over” but it does mean that other tech sectors will find an easier time recruiting talent to their regions and away from Silicon Valley. And talent is really the only thing that matters these days.

6/ Cryptography emerged in the 2010s as a powerful technology that can solve some of the web and mobile’s most vexing issues. Cryptography and encryption have been around for a very long time, well before the computer. Modern computer cryptography came of age in the 1970s. But the emergence of the internet, web, and mobile computing largely did not integrate many of the central ideas of cryptography natively into the protocols that these platforms were built on. The emergence of Bitcoin and decentralized money this decade has shown the way and set the stage for cryptography to be built natively into web and mobile applications and deliver control back to users. Credit to Muneeb Ali for framing this issue for me in a way that makes a lot of sense.

7/ Technology inserted itself right in the middle of society this decade. Our President wakes up and fires off dozens of tweets, possibly while still in bed. We are all hostage to our phones and the services that we rely on. Our elections are conducted using machine learning technology to segment and micro-target important voting groups. And bad actors can and do use the same technologies to interfere in our elections and our public discourse. There is no putting the genie back in the bottle in this regard, but the fact that the tech sector has such a powerful role means that it will be highly regulated by society. And there is no putting the genie back in the bottle in that regard either.

8/ The rich got richer this decade. Axios wrote in a recent email that:

“The rich in already rich countries plus an increasing number of superrich in the developing world … captured an astounding 27% of global growth.”

But the very poor also had a great decade as Axios also reported:

The rate of extreme poverty around the world was cut in half over the past decade (15.7% in 2010 to 7.7% now), and all but eradicated in China.

The losers in the 2010s were lower middle class and middle class people in the developed world whose incomes stagnated or fell.

Technology played a role in all of this. Many of the superrich obtained their wealth through technology business interests. Some of the eradication of extreme poverty is the result of technology as well. And the stagnation of earning power in the lower and middle class is absolutely the result of technology automation, a trend that will only accelerate in coming years.

9/ This a post publish addition. A huge miss in my original post is the emergence of China as a tech superpower and a global superpower. There are many areas (digital money for example) where China is light years ahead of the western world in technology and that will likely accelerate in the coming years. Being a tech superpower is a necessary condition to being a global superpower and China is already that and getting more powerful by the day.

I will end there. These are the big mega-trends I think about when I think about the 2010s. There is no doubt that I left out many important ones. You can and will add them in the comments (wordpress for now), emails to me, and on Twitter and beyond. And that is what I hope you will do.

#crypto#entrepreneurship#machine learning#policy#Politics#VC & Technology#Web/Tech

AIVC

My friend Fraser took a large number of AVC blog posts over the years and trained an AI model on them.

The result is a blog written by a machine.

You can see it here.

One one hand, it is kind of amazing that you can train a machine to write like someone.

On the other hand, I don’t think I will be out of a job anytime soon.

#machine learning